Basics of Radiography Testing

and Image Processing

B.Venkatraman*, Uwe Ewert**, M.Menaka* and
Uwe Zscherpel ** Uwe Zscherpel **
*Indira Gandhi Centre for Atomic Research , India ,
** BAM, Berlin,Germany , , y
Preconference Workshop 18 WCNDT, Durban, April 14, 2012
Structure of the Talk
Basic Principle of Radiographic Testing
Image formation in radiography and Image formation in radiography and
factors affecting radiographic image
N d f I i Need for Image processing
Image parameters and terminologies age pa a ete s a d te o og es
Image Processing Methods
Applications on radiographic images
Case Study Case Study
Radiography Radiography
116 yrs old Technique is identical.
Improvements in sources and detectors.
Object placed between
source and detector (film).
Differential absorption of
radiation. I = I
o
e
-x
o
The film darkness
(density) will vary with (density) will vary with
the amount of radiation
reaching the film
X-ray film
= less exposure
g
through the test object.
= more exposure
p
Top view of developed film
Sources Detectors
Conventional (kV mA)
Film (D2 D7) Film (D2 D7)
Conventional (kV, mA)
Microfocus
Nanofocus
Film (D2..D7) Film (D2..D7)
Image intensifiers Image intensifiers
Nanofocus
Linac
B t t
XX--ray sensitive ray sensitive vidicons vidicons
Fluorescent screens + Fluorescent screens +
Betatron
Synchrotron
CCD CCD
Imaging Plates Imaging Plates
Isotopic Sources
Ir
192
, Co
60
, Tm
170,
Imaging Plates Imaging Plates
Digital Flat Panels Digital Flat Panels
Li Di d A Li Di d A
Yb
169
, Cs
137
, Se
75
Linear Diode Array Linear Diode Array
PARAMETERS AFFECTING RADIOGRAPHIC IMAGING
MEASURED THROUGH SENSITIVITY
SUBJECT
CONTRAST
FILM CONTRAST*
DEFINITION
CONTRAST
ENERGY
IMAGE QUALITY
INTENSITY
SCATTERED RADIATION
SIGNAL TO NOISE RATIO
NORMALISED CONTRAST
SCATTERED RADIATION
FOCAL SPOT SIZE
NORMALISED CONTRAST
BASIC SPATIAL RESOLUTION
DETECTOR CHARACTERISTICS
TRADITIONAL APPROACHES - COMPENSATION PRINCIPLES
CONFIGUARATION OF IMAGE PROCESSING SYSTEM
X-RAYS TRANSMITTED THR. OBJECTS
FILMS
REAL TIME SYSTEMS
DIGITAL
DETECTORS DETECTORS
DIGITISER
(SCANNER, A/D
A/D CONVERTOR,
(LUT)
DIGITAL
INTERFACE
CONVERTOR, LUT)
COMPUTER + GRAPHICS CARDS + MEMORIES + DISPLAY DEVICES
DIGITAL RADIOGRAPHIC IMAGE DIGITAL RADIOGRAPHIC IMAGE
DIGITAL IMAGE REPRESENTATION
IMAGES: IMAGES:
Images - 2-D intensity function f(x,y), x and y : spatial coordinates and f :
proportional to the intensity / brightness (gray level) of the image.
Image Analog Digital Image necessitates obtaining finite number of Image Analog Digital Image necessitates obtaining finite number of
rows and columns and quantized levels.
DIGITAL IMAGE:
Sampling
Image
Sampling
(x,y)
Quantizatio
Digital
Image
Finite rows and Quantizatio
n f(x,y)
Finite rows and
columns and finite
gray levels
It is a matrix with M rows and N columns. Each matrix element has a
typical gray level value ranging from 0-255 (8 bit).
Colour Image three functions can be written as a vector valued function Colour Image three functions can be written as a vector valued function
( , ) r x y


( , ) ( , )
( , )
f x y g x y
b x y

IMAGE FORMATS AND TYPES
Converting the acquired image into computer acceptable
digitized from.
Image can be of any of stored formats and image types Image can be of any of stored formats and image types.
Stored formats
TIFF (Tagged Image File Format)
JPEG (Joint Photographic Expert Group)
BMP (Windows Bitmap) BMP (Windows Bitmap)
XWD (X Windows Dump)
PNG (Portable Network Graphics)
GIF (Graphics Interchange Format) GIF (Graphics Interchange Format)
Image types
Intensity image (data matrix with intensities scaled to represent gray
scale)
Bi I (L i l f 0 d 1 ( l bl k d hit )) Binary Image (Logical array of 0s and 1s (only black and white))
RGB image( three separate matrices (Red, Green and Blue)
EFFECT OF NUMBER OF PIXELS
16 X16 32 X 32
64 X 64 256 X 256
512 X 512
1024 X 1024
EFFECT OF GREY LEVELS
1 bit 2 grey values 1 bit -2 grey values
2 bit -4 grey values 3 bit -8 grey values 5 bit -32 grey values
6 bit -64 grey values 8 bit -256 grey values
SPATIAL & GRAY LEVEL RESOLUTION
Image Resolution
Sum total of Pixel Resolution and Grey Level Resolution
Number of Samples determines spatial resolution of an image.
N b f l l d t i G l l l ti f i Number of gray levels determines Gray level resolution of an image.
Subsampling accomplished by deleting appropriate number of rows and p g p y g pp p
columns but allowable gray levels maintained as 256.
Similarly, pixel replication brings all sub sampled images back to original size
f 512* 512 of 512* 512.
Drawback : Checker board pattern will be more pronounced in 64 * 64 and
32* 32 images respectively 32 32 images respectively.
Drawback : False Contouring will be more pronounced in 16 or less uniformly
spaced gray levels spaced gray levels.
IMAGE PROCESSING
Image processing refers to a set of
operations performed to transform an operations performed to transform an
existing image f into new image g.
We can transform either the range of f We can transform either the range of f.
Or the domain of f: Or the domain of f:
Image Processing - Objective
Primary purpose of image processing in radiology includes
Improvement image quality making it more clear to
human perception : improve sensitivity and contrast.
Render minute details - finer defects can be identified.
P i Q tit ti f t h t i ti th h Precise Quantitative feature characterisation through
enhancement of edges.
Make data amenable for archiving and transmission.
Render it amenable for automated defect recognition. g
Some argument about where image processing ends
and fields such as image analysis and computer vision and fields such as image analysis and computer vision
start.
DIGITAL IMAGE PROCESSING
Processes whose inputs and outputs are images in which we
render detail more clearly and extract attributes from images up
to and including the recognition of individual objects.
Image Compression (for effective
& )
Morphological Image Processing
( )
to and including the recognition of individual objects.
Wavelets & Multi Resolution
storage & transmission) (image component extraction)
Color Image Processing
( ti i l d i )
Wavelets & Multi Resolution
Processing
Image Segmentation
Knowledge base
Description
(Quantitative
characterization)
Image Restoration
(Recover image)
(operations in color domain)
characterization)
Image Enhancement (High-
light Region of Interest)
(Recover image)
Object Recognition
(assign label to
Image Acquisition
objects)
Image Processing Operations typical in RT
Image Processing
Image enhancement
Neighbourhood
Image enhancement
- Point Operations
- Brightness
Neighbourhood
/Matrix operations
- Smoothing and
Sh
Brightness
- Contrast
- LUT
Sharpness
- statistical filters
(Median Max Min)
- Gamma value,
- Histogram
(Median, Max,Min)
- directional filters
- gradient filter g
- edge detection
P i S ti l D i i d t d b th i
Point Operations
Process in Spatial Domain is denoted by the expression
G (x,y) = T [ F(x,y) ]
F(x,y) is Input Image, G (x,y) is Processed Image
T the operator applied at each location (x y) to yield output G T the operator applied at each location (x,y) to yield output G.
Gray Level Transformation:
S T( ) S =T(r)
The effect is to produce an image of higher contrast than the original by
darkening the values of r below m and brightening the values above m darkening the values of r below m and brightening the values above m .
This results in Contrast Stretching.
Light
s
Light
m
Dark Light
Dark
r
Contrast and Brightness Enhancement
Raw Image
Raw Image After Brightness and Contrast Adjustment
Histogram Equalisation
k
S
k
= T(r
k
) = k = 0,1, 2 ..L 1, r
k
is the kth gray level
n
k
is the number of pixels in the image having gray level r
k

k
j
j n n
0
/
The idea of histogram equalization is that the pixels should be
distributed evenly over the whole intensity range, i.e. the aim is to
3
x 10
5
transformtheimagesothat theoutput imagehasaflat histogram.
1
1.5
2
2.5
0 50 100 150 200 250 300
0
0.5
Original radiograph
2
2.5
3
x 10
5
0 50 100 150 200 250 300
0
0.5
1
1.5
Histogram enhanced
Neighbourhood / Matrix Operations
Z
1
Z
2
Z
3
W
1
W
2
W
3
Response R of a pixel is
Z
4
Z
5
Z
6
W
1
W
2
W
3
W
4
W
5
W
6
R= w
i
z
i
i=1, 2, ., mn
Window is moved in such a way
Z
7
Z
8
Z
9
W
7
W
8
W
9
Image neighborhood
that W
5
is placed on the pixel of
interest.
Image divided into neighborhoods.
Image neighborhood
(m=n=3)
Filter or window
(m=n=3)
Image divided into neighborhoods.
Sub images called filter or mask or kernel or template or window is applied
on the neighborhood of the images.
Matrix Operations - Image Enhancement
Image Enhancement
Refers to highlighting region of interest in thermogram
Spatial domain analysis
Frequency Domain analysis
Spatial domain analysis
(filter masks are applied on
thermograms directly in
spatial domain)
q y y
(converting thermogram into
frequency domain and then filter
masks are applied)
p )
Smoothing
Sharpening
(Highlights fine detail in thermogram)
(Highlights a portion of thermogram by
blurring to remove noise)
Li (A i G i filt ) & N Li
(Highlights fine detail in thermogram)
Second order (Laplacian) & First order
(Sobel, Prewitt and Roberts filters)
Linear (Averaging, Gaussian filters) & Non Linear
filters (Median, maximum and minimum filters)
(Spatial domain)
Butterworth, Gaussian and ideal Low pass filters
(Spatial domain filters)
Butterworth, Gaussian and ideal High
pass filters (Frequency domain)
, p
(Frequency domain)
pass filters (Frequency domain)
Smoothing
Original Image
P(i,j) = average [P(i,j), P(i-1,j-1), P(i+1,j+1),P(i+1,j-1), P(i-1,j+1)]
Smooth : Blurs the active
image or selection. This filter
l h i l ith th replaces each pixel with the
average of its 33
neighborhood.
Smoothened Image
Averaging Filters
Averaging filters
Replace the gray level of the candidate
cropped image
pixel by the average gray level of the
neighborhood pixels.
Eliminate impulse noise and
d d ti i thi t i degradations appearing as thin stripes.
Blur the image to get a gross
representation of objects of interest as
small deviations due to noise blend with
Fig. a Original Radiograph
small deviations due to noise blend with
the background while objects of interest
are easy to detect.
Enhanced radiograph using average filter
1 1 1
1 1 1
1/9
1 2 1
2 4 2
1/16
Fig b Average filtered Radiograph
1 1 1
Averaging Mask
1 2 1
Weighted Averaging
Mask
Fig. b Average filtered Radiograph
Sharpening Filters - Spatial Domain
To highlight fine detail in an image or to enhance detail that has
been blurred either in error or during image acquisition
Fine details result in discontinuities in an image such as points,
edges and lines
Operator used must result in zero values at flat segments and
non zero at discontinuities. non zero at discontinuities.
First order derivative (Gradient operators)
Mathematical equation : f/ x=f(x+1,y)-f(x,y), f/ y=f(x,y+1)-f(x,y)
Filter masks : Roberts, Sobel and Prewitt
Second order derivative (Laplacian operators) Second order derivative (Laplacian operators)
Mathematical equation :
2
f/ x
2
=f(x+1,y)+f(x-1,y)-2f(x,y), Mathematical equation : f/ x f(x 1,y) f(x 1,y) 2f(x,y),

2
f/ y
2
=f(x,y+1)+f(x,y-1)-2f(x,y)
Filter mask : Laplacian
Filtering Noise Removal
Mean Filter (Matrix mean) Median Filter Mean Filter (Matrix-mean) Median Filter
Edge Enhancement
Convolve filter Variance filter Band Pass Filter Unsharp mask
High Pass Filter
Details of the Operations Performed on the
Images
Media Filter: Reduces noise in the active image by replacing each pixel with the median of
the neighboring pixel values
.
g
Mean Filter: Smooths the current image by replacing each pixel with the neighborhood
mean.
Minimum Filter: Filter does grayscale erosion by replacing each pixel in the image with the g y y p g p g
smallest pixel value in that pixels neighborhood.
Maximum Fitler: Filter does grayscale dilation by replacing each pixel in the image with the
largest pixel value in that pixels neighborhood. a gest p e a ue t at p e s e g bo ood
Gaussian Blur: Filter uses convolution with a Gaussian function for smoothing
C l D ti l l ti i k l t d Convolve: Does spatial convolution using a kernel entered
Unsharp Mask: Sharpens and enhances edges by subtracting a blurred version of the
image (the unsharp mask) from the original.
Variance Filter: Highlights edges in the image by replacing each pixel with the
neighborhood variance.
Bandpass Filter: Removes high spatial frequencies (blurring the image) and low spatial
frequencies (similar to subtracting a blurred image). It can also suppress horizontal or
vertical stripes that were created by scanning an image line by line.
Case Study Case Study - -1 1
In In- -service Inspection service Inspection Power Plant Life Extension Power Plant Life Extension pp
Sodium blockage in a hot argon communication
line of Fast Breeder Test Reactor at Kalpakkam. p
Reactor shut down. Detection became critical.
Low density material. Gamma source Ir 192. Low density material. Gamma source Ir 192.
Challenge Limited access, large area, low Z
material Very low contrast radiograph material. Very low contrast radiograph.
Approach - Advanced Image processing :
Contrast Stretching Noise Removal Contrast Stretching, Noise Removal,
Sharpening - Edge Enhancement
Case Study 2 Enhancement of Tenon Rivet Head
in HP Rotor of MAPS Unit -1 Turbine Generator in HP Rotor of MAPS Unit 1 Turbine Generator
Power Station in India shutdown (as part of bi-
annual maintenance) and Turbine Generator was )
opened for in-service inspection. During in-service
visual inspection, the thickness of rivet head of
tenon of blade 17 of HP rotor stage-2 was found to g
be less than other blade tenons and almost flush
with shroud. Regulators wanted proof.
Low Contrast radiograph by Gamma Radiography
Approach Denoising
+ Deblurring +
Gaussian Filter + Gaussian Filter +
bandpass filter +
contrast enhancement
and profiling
Filtered and contrast enhanced image. Note that the
edges of the rivet head can be seen as sharp points over
the shroud portion.
and profiling
Contrast &
brightness
Histogram
Flow chart
developed
for Image
Equilisation
g
processing
Edge Enhancement through Spatial filtering
Case Study 3 : Quality Assurance of
small diameter 0.5 mm Thermocouple
ASTM Standard + Image Enhancement =
E h d I t t ti d Q tit ti
small diameter 0.5 mm Thermocouple
used in nuclear reactors
Enhanced Interpretation and Quantitative
Evaluation for
weld quality, sheath spacing, uniformity of
insulation material filling insulation material filling
Materials Evaluation, May 2007
COMBINING ADVANCED CONCEPTS WITH
STANDARD PRACTICES FOR ROBUST
QUALITY ASSURANCE
Thank You Thank You